Horton

Re: logical replication: restart_lsn can go backwards (and more), seems broken since 9.4

Tomas Vondra <tomas@vondra.me>

View thread View on postgr.es
From: Tomas Vondra <tomas@vondra.me>
To: Masahiko Sawada <sawada.mshk@gmail.com>
Cc: Amit Kapila <amit.kapila16@gmail.com>, Ashutosh Bapat <ashutosh.bapat.oss@gmail.com>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2024-11-19T21:36:08Z
Lists: pgsql-hackers

Attachments

OK,

Now that the fix is committed, shall we introduce some of the asserts? I
believe there's an agreement the restart_lsn shouldn't move backwards,
so I propose the attached patch.

I still think not resetting the fields when releasing the slot, and
allowing the values to move backwards is rather suspicious. But I don't
have any reproducer demonstrating an issue (beyond just hitting an
assert). Perhaps it's correct, but in that case it'd be good to add a
comment explaining why that's the case. Sadly, it has to be written by
someone else - I've been unable to form a justification why it's OK :-(

regards

-- 
Tomas Vondra

Commits

  1. Fix a possibility of logical replication slot's restart_lsn going backwards.

  2. Introduce logical decoding.